Detecting and understanding crash-consistency bugs across the parallel I/O stack

Parallel file systems (PFS’es) and parallel I/O libraries have been the backbone of high- performance computing (HPC) infrastructures for decades. However, their crash consistency bugs have not been extensively studied, and bug-finding or testing tools for identifying them are lacking. In this dissertation, we developed a generic testing system PFSCheck to study crash consistency bugs on popular PFS’es, such as BeeGFS and OrangeFS, with a cross-stack approach that covers HPC I/O library, PFS, and their interactions with local file systems. We evaluated the crash vulnerabilities of common HPC workloads and the evaluation results show that PFS’es and I/O libraries suffer from more crash consistency bugs than regular file systems due to the scale and complexity of the I/O stack. Local file system configurations such as journaling mode and the choice of consistency checker also influence the discovered vulnerability pattern.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2020

Author dc:creator

• Sun, Jinghan

Contributors dc:contributor

• Snir, Marc

• Parallel File System
• Crash Consistency
• Bug identification
• High Performance Computing

Statement dc:rights

• Copyright 2020 Jinghan Sun

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/108145

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/108145